Add absolute value to montecarlo plot

Author: pachoning

survlimepy/survlime_explainer.py

@@ -394,6 +394,7 @@ def plot_montecarlo_weights(
         scale_with_data_point: bool = False,
         figure_path: Optional[str] = None,
         with_colour: bool = True,
+        absolute_vals: bool = False,
     ) -> None:
         """Generates explanations for a prediction.
 
@@ -403,6 +404,7 @@ def plot_montecarlo_weights(
             scale_with_data_point (bool): whether to perform the elementwise multiplication between the point to be explained and the coefficients.
             figure_path (Optional[str]): path to save the figure.
             with_colour (bool): boolean indicating whether the colour palette for positive coefficients is different than thecolour palette for negative coefficients. Default is set to True.
+            absolute_vals (bool): whether to plot the absolute values of the coefficients.
 
         Returns:
             None.
@@ -454,9 +456,25 @@ def plot_montecarlo_weights(
         colors_up = {key: val for key, val in zip(median_up.keys(), pal_up)}
         colors_down = {key: val for key, val in zip(median_down.keys(), pal_down)}
         custom_pal = {**colors_up, **colors_down}
-        data_reindex = data.reindex(columns=custom_pal.keys())
-        data_melt = pd.melt(data_reindex)
-
+        if absolute_vals:
+            absolute_order = {**median_up, **median_down}
+            absolute_order = {key: np.abs(val) for key, val in absolute_order.items()}
+            absolute_order = dict(
+                sorted(
+                    absolute_order.items(),
+                    key=lambda item: np.abs(item[1]),
+                    reverse=True,
+                )
+            )
+            custom_pal = {key: custom_pal[key] for key in absolute_order.keys()}
+            data_reindex = data.reindex(columns=list(custom_pal.keys()))
+            data_melt = pd.melt(data_reindex)
+            data_melt.value = np.abs(data_melt.value)
+            plot_title = "Absolute feature importance"
+        else:
+            data_reindex = data.reindex(columns=custom_pal.keys())
+            data_melt = pd.melt(data_reindex)
+            plot_title = "Feature importance"
         _, ax = plt.subplots(figsize=figsize)
         ax.tick_params(labelrotation=90)
         if with_colour:
@@ -481,7 +499,7 @@ def plot_montecarlo_weights(
         p.yaxis.grid(True)
         p.xaxis.grid(True)
 
-        p.set_title("Feature importance", fontsize=16, fontweight="bold")
+        p.set_title(plot_title, fontsize=16, fontweight="bold")
 
         plt.xticks(fontsize=16, rotation=90)
         plt.yticks(fontsize=14, rotation=0)